Skip to main content

es 数据恢复操作步骤

import BrowserWindow from '@site/src/components/BrowserWindow';

1. 推迟分片分配​

默认情况，集群会等待一分钟来查看节点是否会重新加入，如果这个节点在此期间重新加入，重新加入的节点会保持其现有的分片数据，不会触发新的分片分配。 通过修改参数 delayed_timeout ，默认等待时间可以全局设置也可以在索引级别进行修改:

```console PUT _all/_settings { "settings": { "index.unassigned.node_left.delayed_timeout": "5m" } } ```

2. 禁用副本分配​

停止外部写入，停止副本分配

```console PUT _cluster/settings { "persistent": { "cluster.routing.allocation.enable": "none" } }

"cluster.routing.rebalance.enable": "none",

"cluster.routing.allocation.allow_rebalance": "indices_primaries_active"

</BrowserWindow>

## 3. 执行同步刷新

<BrowserWindow>
```console
POST _flush/synced

4. 停止服务​

停止 目标节点上的es服务

```shell sudo systemctl stop elasticsearch.service # sudo service elasticsearch stop ```

5. 还原配置​

等主分片恢复后，还原配置

```console

可省略

PUT _cluster/settings { "persistent": { "cluster.routing.allocation.enable": "primaries" } }

"cluster.routing.rebalance.enable": "primaries",

"cluster.routing.allocation.allow_rebalance": "indices_primaries_active"

PUT _cluster/settings { "persistent": { "cluster.routing.rebalance.enable": "all" } }

"cluster.routing.allocation.enable": "all",

"cluster.routing.allocation.allow_rebalance": "indices_all_active"

</BrowserWindow>




## 其他

### 手动分配分片到节点

<BrowserWindow>

POST _cluster/reroute { "commands":[ { "allocate_stale_primary":{ "index": "anxinyun_aggregation", "shard": 1, "node": "es-n4", "accept_data_loss" : true } } ] }

</BrowserWindow>

Last updated on